Wednesday, March 11, 2009

Improved Quality - Silver Linings and Other Truths

Every cloud has a silver lining, right? Sometimes it's not so easy to see it when it is raining on your parade, but I've found the silver lining to be mostly true. You just have to step back and take a look.

In the current case, I got a "bad" batch of pipe from which I made a number of whistles. It wasn't really that bad, not as far as water pipe goes. But it was bad as far as my whistle making was concerned. I had been making the Every family of whistles the same way for quite a while with no problems and had relaxed my inspection from 100% to only several per batch, following the tried and true QA US Govt Mil Spec sampling plans. All was going along fine. 

After sending whistles to a few really experienced whistle players with good ears, I got a phone call from two, and an e-mail from the other saying the intonation of the whistle was off. And not just a little bit. Hmm... what could it be? Could you send me a clip of yourself playing the whistle, and then return the whistle to me for evaluation? I wanted to see how the whistles checked out in my hands compared to theirs. And it looks like I'm back to 100% inspection againg. 

Could it just be due to different playing styles? None of the whistles I put thru the stringent testing showed a problem here in my hands. These folks have all played a number of different whistles so probably not. I'm expecting a problem with the whistles.

I had recently begun using the RTTA intonation analysis software from Terry McGee and Graeme Roxbury  and had been testing some whistles with it, especially if I was changing the location of finger holes or something. I took the clips and ran them through the RTTA software. I now had a pretty good picture of the "live" intonation of the whistles in the customer's hands. When I got them back and did a plot myself, my plots showed intonation problems as well. So I reported that fact to the folks who were kind enough to tell me about it and set to work on finding out why these whistles were so different from the others I have made. 

From Whistle Promos
Poor intonation whistle example. Only a handful of the whistles were this bad due to the thickness of the pipe from the maker. But it gives a good idea of how much a little extra material in the wrong place can change things. (click to enlarge)

My first reaction was to think that there must be something I did wrong. Maybe I got the calipers calibrated to something other that zero for a while. Something. But everything thing I normally measure checked out fine. 

So I had to start checking things I didn't usually measure. The problem turned out to be a narrower body ID for the whistles. It wasn't much when measured, less than one-half of a millimeter, but in relation to the other dimensions that's quite a lot, as it applied for the whole length of the whistle. 

The pipe, when used as hot water pipe, has to hold a minimum pressure, which requires a minimum wall thickness. The pipe must also fit properly into fittings, so the OD of the pipe is the most critical dimension for the plumbers. So if the pipe comes out a little heavy, it still makes good pipe, it just costs the maker some extra material. So the maker wants to use the minimum material, which causes the ID to be pretty darn consistent. 

Until now. Maybe they started up a new line or maybe they started up a new operator. Whateve the cause, some of the pipe had a smalled ID than I had been getting from this source for years. OK, now that I know what happened, what can be done about it? 

I could just measure the pipe and try to use only that pipe which was the size I expected it to be. But just beacuse I measure one spot doesn't mean the whole pipe is the same size. So I decided to take matters into my own hands and ream the ID of all the body sections so I would be sure that the body ID was the same for all my whistles. 

I made the prototype whistles, and found there was an improvement in the intonation right off the drill press, which required less work tuning the whistle. This was good since reaming adds a fair amount of time to the making of each whistle and I hoped to get that time back somewhere. Of course happier customers alone are worth it, but often there's an actual manufacturing savings when you improve product quality. So long as you actually improve the process and not just inspect the quality in by throwing out more product. 

Now that I was going to be reaming the pipe, I was going to be chucking the pipe quite tight and would mar the end of it with the chuck teeth. So this would need to be cut off, wasting a small section of material. Since I was now reaming the body I thought it would make sense for me to do all the body parts of a whistle together and keep them as a unit now rather than make so many heads, so many bodies and so many Walkabout parts and then after they are all machined, match them together. This required a complete change in how I measured the cuts when parting the whistle pieces from the original pipe. 

Well, you get the idea. One small improvement and I completely re-worked my process for making the whistles. They appear to be the same whistles when you pick them up. But from my perspective I make them in a very different way now. 

From Whistle Promos
This is a plot from one of the whistles made after the process change. (click to enlarge)

Even before the process change I have been keeping statistical process control charts on a lot of the steps during the making of a batch of whistles. Here's one on the tuning slide outside diameter. You can see just from looking that they are getting more consistent over time, due to relatively minor changes in how I do things. Note that it's how I do things not what I do that is improving. Maybe I'll talk more about SPC in the future.

From Whistle Promos
In a nutshell, narrower limits on the top chart means more uniform part. (click to enlarge)